Probability of a cyclone beginning rapid intensification was greatest in the Gulf of Mexico and lowest in the northeast Atlantic. It did not display a peak during the early September peak of overall cyclone occurrence. Cyclones were most likely to begin rapidly intensifying immediately after cyclogenesis, and at that time were most likely to do so if they displayed 35 kt initial wind speeds. Cyclones of > 1 d age showed peaks of intensification onset at 30 kt and 70 kt. Intensification was most likely to begin when winds were unchanged over the preceding 6 h, and when central pressure had dropped by 1 or 2 mb over 6 h. In three time zones, intensification of 15 kt/24 h appeared most likely to begin shortly after midnight, and least likely shortly before midnight. I discuss the possibility that spurious patterns could have resulted from overlapping influences of explanatory variables, or from biases in procedures of storm observation.